Technology’s underlying systems had an unusually difficult week. A leadership crisis hit Nvidia, Microsoft 365 suffered a global outage, and security teams received urgent new reasons to patch internet-facing systems.
Elsewhere, financial results and health-tech developments showed that the questions around infrastructure are not only technical. They are also about who pays, who is accountable, and who gets to set the rules.
Nvidia puts its CFO in the CEO role
Jensen Huang has resigned as Nvidia’s CEO, president, and board member, according to Inc. The departure followed reports of relationships that could create conflicts of interest. Nvidia CFO Colette Kress will serve as interim CEO while continuing as CFO.
The board has retained Wachtell, Lipton, Rosen & Katz for an internal investigation. Nvidia says its operations, customer commitments, and product roadmap are expected to continue without interruption.
Kress gives the company an experienced internal operator at a moment when continuity matters. But an interim CEO who also remains CFO is necessarily a temporary arrangement. Nvidia’s customers and partners will be watching for two things: what the investigation finds, and whether the board can run a credible search for a permanent successor without distracting the company from delivery.
Patch Apache HTTP/2, then review the rest of the stack
Apache HTTP Server fixed CVE-2026-119618, a critical remote-code-execution flaw in its HTTP/2 module. The issue affects versions 2.4.55 through 2.4.70, and version 2.4.71 includes the fix. The Hacker News reports that unauthenticated attackers could remotely execute code.
For administrators, this is a straightforward priority: identify exposed Apache servers, confirm whether HTTP/2 is in use, and move to the fixed version or apply available mitigations. A remote code-execution flaw on a public web server is not a ticket to leave in the normal patch queue.
The same advisory cycle also included 16 Node.js flaws, including HTTP/2 issues that can crash servers or expose data from other connections. It also included a pre-authentication MongoDB issue that can crash lagging servers during failover, plus claims of additional N-central exploits against patched on-premises servers. The useful lesson is not to treat a prominent Apache alert as the whole story. Teams should review the components that actually run their services.
Microsoft traces a 365 outage to a network upgrade
Microsoft attributed a major global Microsoft 365 outage to a network upgrade, according to Windows Central. Outlook, Teams, OneDrive, and Copilot were affected for more than 18 hours, and Microsoft apologized after roughly 24 hours of disruption.
That scope is a reminder of how much ordinary business work now sits behind one cloud login. Communication, documents, meetings, and files may be different applications to users, but a platform-level failure can take them down together.
Email also has an awkward recovery pattern. As The Conversation U.S. explains, restoring the original service does not instantly restore normal delivery. External mail systems can keep retrying messages, creating a backlog that takes time to clear. Companies that depend on Microsoft 365 should treat an alternate communications channel and a realistic outage plan as operational basics, not as disaster-recovery paperwork.
Cisco sees the network side of data-center spending
Cisco reported fiscal first-quarter revenue of $18.5 billion, up 24% year over year. Its networking revenue rose 43% to $11.1 billion. The company also raised its fiscal 2027 forecast for hyperscaler AI-infrastructure revenue to $9 billion, more than twice its stated fiscal 2026 total.
Those figures matter because data-center expansion is not only a chip story. Large computing deployments need networks to connect servers, storage, and services. Cisco’s results suggest established networking suppliers are benefiting substantially from the same buildout.
The growth is uneven, though. Cisco said networking is expanding faster than its security, collaboration, and services units. It also completed the acquisition of WideField Security during the quarter. For buyers and investors, the important detail is the mix: the infrastructure spending surge is concentrating in particular parts of the technology stack.
BCI trial participants ask for a voice after the trial ends
The BCI Pioneers Coalition, a group of people participating in brain-computer-interface trials, has published preliminary participant-led guidelines for researchers and companies. The recommendations cover compensation, recognition for research contributions, access to neural data, limits on restrictive nondisclosure agreements, media privacy, and plans for device maintenance or removal after trials end.
There are few people with such implants, so the guidelines draw on a small group. That limitation is real, but it also explains why their experience should carry weight. These participants understand the practical consequences of trial design in a way that product teams and ethics boards may not.
The most important request is a mundane one: plan for what happens after funding or a study ends. An implant can become useful to a participant, yet a research project may have no built-in commitment to maintain it. This looks like an early effort to establish that neural data and implanted hardware deserve obligations beyond a conventional software trial.
CoreWeave’s growth comes with a large financing bill
AI cloud provider CoreWeave reported third-quarter revenue of $3.66 billion, up 168% year over year. It also reported a $709 million net loss and $846 million in quarterly interest expense.
The company expects $9 billion in customer prepayments and deposits over the next three months. It plans to use loans tied to specific customer contracts to keep expanding. CoreWeave is also favoring shorter, higher-priced capacity contracts, which offer lenders less predictable cash flows.
The numbers capture the uncomfortable mechanics of supplying high-end computing capacity. Demand may be strong, but building and financing the capacity is expensive. Customer prepayments can reduce some pressure, yet they also make customers part of the financing structure. That is a model worth understanding for anyone buying long-term compute capacity from a fast-growing provider.
Oura tests a path from wellness data to clinical reimbursement
Oura plans to acquire AI doctor startup Counsel Health for $625 million, according to the Mario | Health Tech newsletter. Counsel had previously disclosed raising $36 million. The companies already work together in Medicare’s ACCESS model, which tests outcome-linked payments for chronic-disease care.
The deal is an attempt to connect consumer wearable data with services that can be paid for as clinical care. That is a much harder business than selling a device and subscription. Medicare’s model and insurers’ willingness to reimburse care will help determine whether the approach can grow.
The companies’ links to Eli Lilly also invite scrutiny. Counsel has a GLP-1 referral program, while Lilly has invested in Oura. Oura’s CEO disputes concerns about incentives, but the underlying question is reasonable: when health data, medical recommendations, and commercial referrals meet in one service, the relationships need to be clear.
AlphaGenome 2 narrows genetic predictions to cell context
Google DeepMind released AlphaGenome 2, a research model designed to predict how DNA variants affect gene regulation in specific tissues, cell types, and cell states. It takes a one-million-letter DNA sequence plus a biological condition as input and draws on more than 100 million single-cell profiles from 68 human and mouse tissues.
DeepMind reports improved benchmark results. The useful scientific idea is that a DNA variant may have different effects depending on where and under what biological conditions it is examined. A model that accounts for cell context could help researchers prioritize variants and pathways for laboratory work.
DeepMind also states the boundary clearly: AlphaGenome 2 is for research and is not validated for clinical or diagnostic use. That distinction should not get lost in the excitement around biological models. A tool that helps select experiments is not a system for interpreting an individual patient’s genetic test.
CISA flags active exploitation of UnrealIRCd and Check Point flaws
CISA added two vulnerabilities to its Known Exploited Vulnerabilities catalog: CVE-2026-114713 in UnrealIRCd and CVE-2026-123039 affecting multiple Check Point products. CISA cited evidence that both are being actively exploited.
The issues are described as improper delimiter neutralization in UnrealIRCd and incorrect permissions for a critical resource in affected Check Point products. Federal civilian agencies must prioritize remediation under Binding Operational Directive 26-04.
A KEV listing is not merely another item in a vulnerability feed. It is a practical signal that patch-management teams should identify exposed systems and move remediation up the list. Private-sector organizations are not subject to that federal directive, but the active-exploitation evidence is relevant to them too.
OpenAI funding report needs a careful reading
Bloomberg, citing people familiar with the matter, reportedly said that OpenAI raised $11 billion toward a funding round expected to total about $20 billion at a $1.25 trillion valuation. The round had not closed, and OpenAI had not commented, according to Supercharged With AI.
The report says OpenAI had previously been reported to be seeking up to $30 billion. If the financing closes, it would provide substantial capital for computing infrastructure and product development. But this is precisely the kind of story where the conditional language matters. The final amount, valuation, and terms should be treated as unconfirmed until the company or investors disclose them.
Sources
- Inc.
- The Hacker News
- Windows Central
- The Conversation U.S.
- Cisco Systems Inc.
- The Checkup from MIT Technology Review
- Supercharged With AI
- Mario | Health Tech
- The Neuron
- CISA